何毓琦的个人博客分享 http://blog.sciencenet.cn/u/何毓琦 哈佛(1961-2001) 清华(2001-date)

博文

On Artificial Intelligence 精选

已有 5883 次阅读 2018-12-6 22:32 |个人分类:S and T|系统分类:海外观察

For new readers and those who request to be “好友 good friends” please read my 公告 first


Recently because of the news concerning Alpha Go, Chess, the championship of the game Jeopardy, and most recently, AlphaFold (http://blog.sciencenet.cn/blog-927304-1150143.html)  et al, the subject of AI has reached the level of general interest among the public consciousness. My comments below are from the viewpoints of an interested and informed public and not from that of experts working in the field.

This much is clear

1.   For problems in specific domain with well defined rules of operation however complex, such as chess or protein folding, computers will be replacing humans and take over increasingly with superior performances.

2.   There will be the social problem of permanently displaced workers in those domains.

Before diving into the sciences and technologies of AI, let us briefly examine some social issues which concern the general public.

First before we let AI to take over tasks that we human used to perform we need to distinguish between “trust and consequences”. In case of failure of the AI, the consequences behind Alpha Go and that of self- driving cars have vastly different importance. Similarly, applications of AI in national security and commercial situations entail different constraints.

This leads to the second point of “weak vs. strong AI”. Weak AI really means IA (Intelligent Assistance). We let computers do what they do best, e.g. , vast memory and fast computing speed, and reserve the final or high level decision to humans. On the other hand Strong AI implies letting computers to take over completely such as in the case of AlphaGo where except defining the rules of the game, the computer takes over data generation, training, learning, and optimization completely.

Thirdly, regardless of weak or strong AI, we always need human leaders with knowledge and authority over the use of AI. Are we training enough personnel with such appreciation and knowledge? To paraphrase an example. Everyone understands the power of nuclear weapons. But authority of its uses should be entrusted only to leader with appreciation of their use in war, foreign policy, and societal responsibilities.

Now we turn to the technology behind AI. “Learning” and “Deep Learning” are words that are fundamental in AI. “Learning” is nothing but automating what we humans do – i.e., iterate on the “observe-postulate-test- revise” cycle of building up usable knowledge. To do this, we need Data (observation), Pattern Recognition (postulate a set of rules), Optimization (test-revise) and Long term multi-level Optimization (deep learning). I have previously discussed/written about the science and technology behind each of these subtopics. Interested reader can spent some time looking for them among my past blogs.

Finally, in a future world of interacting agents of AI, we can speculate on the future of human civilization – upheavals and disruption in society, and unintended consequences and danger of massive AI. Two earlier blog articles – Martin Rees’ book http://blog.sciencenet.cn/blog-1565-1139713.html (are human beings just a transitional phase?) and warnings from http://blog.sciencenet.cn/blog-1565-1135784.html , http://blog.sciencenet.cn/blog-1565-1116213.html are just my own musings on this subject.

One more thing, subjects related to the development of AI are game theory, and data science which I have also touched on in the past.

Game Theory G 1- G5

G1. What is Mathematical Game Theory? http://blog.sciencenet.cn/blog-1565-17894.html

G2. What is Mathematical Game Theory (#2) http://blog.sciencenet.cn/home.php?mod=space&uid=1565&do=blog&id=20337

G3. What is Mathematical Game Theory (#3) http://blog.sciencenet.cn/blog-1565-21210.html

G4. What is Mathematical Game Theory ?( #4) http://blog.sciencenet.cn/home.php?mod=space&uid=1565&do=blog&id=21735

G5. What is Mathematical Game Theory ?( #5) http://blog.sciencenet.cn/home.php?mod=space&uid=1565&do=blog&id=21891

G6. Mathematical Game Theory - Recent developments http://blog.sciencenet.cn/blog-1565-919748.html

 

Data sciences D1-D4

Big Data and Data Science http://blog.sciencenet.cn/blog-1565-763330.html

Big Data and Data Science(II) http://blog.sciencenet.cn/home.php?mod=space&uid=1565&do=blog&id=772512

Big Data and Data Science (3)          

Big Data and Data Science (4)

To  quote a well known and ages old phrase. This is an “Endless Frontier  






https://blog.sciencenet.cn/blog-1565-1150245.html

上一篇:How to sell your ideas
下一篇:Professor Doris Tsao, MacArthur Fellowship recipient 2018
收藏 IP: 74.104.185.*| 热度|

11 白龙亮 刘钢 李剑超 强涛 李春来 许培扬 文端智 张士伟 李毅伟 李天成 肖世昌

该博文允许注册用户评论 请点击登录 评论 (2 个评论)

数据加载中...
扫一扫,分享此博文

Archiver|手机版|科学网 ( 京ICP备07017567号-12 )

GMT+8, 2024-12-23 23:55

Powered by ScienceNet.cn

Copyright © 2007- 中国科学报社

返回顶部